1. Description of the study site
The study was conducted in 3 areas that have both not-for-profit MH+ and MH- health centres: see map (the latter was constructed by the first author, with the help of a cartographer, using the ArcGIS software available for free on line).
Insert map
2. Recruitment of participants
12 health centres were selected among not-for-profit health facilities with similar administrative status, funding patterns and operations. Four of these 12 facilities had integrated a mental health package in their routine offer, eight had not. We chose not to include public or private-for-profit facilities in our study because they had either not integrated mental health in their package of activities or had an organizational culture and a way of functioning substantially different from the selected facilities.
Most of these 12 selected health centres are run by doctors and they are integrated in the local district health systems. In these 12 health centres, we included all providers in charge of the curative consultations in the out-patient department on a daily basis: eventually, this added up to 18 health workers (7 in MH+ facilities of which 5 doctors, 1 nurse-practitioner and 1 social worker; and 11 in MH- facilities of which 9 doctors and 2 nurse-practitioners). Our sample size (450 patients) was calculated to demonstrate a 50% satisfaction rate among participants about their involvement in the consultation process, with a precision of 5%. A sample of at least 175 inclusions in each group allows to detect a difference of 15% (at alpha 5% and beta 80%). In order to have a balanced representation per care provider, we opted to include an equal number, namely 25, of consultations per care provider resulting in 175 consultations for MH+ and 275 for MH- group (ratio MH-/MH+ of 1.5).
Data collection and entry
3. Description of the data collection tool
The PPS tool (9) has been used in various studies to assess patient participation (31,32). It includes 6 questions that probe the extent to which patients feel that they have been able to effectively participate in the decision-making process about their health problem and treatment. The tool was adapted so as to facilitate understanding by study participants. More specifically, three questions (i.e. Q7-9) were added to the PPS tool to assess overall satisfaction and the patient’s preferred level of involvement in treatment decision-making (active or passive). The adapted tool was then translated into the two main local languages (Soussou and Poular; the Malinké language being only very rarely spoken) through a 3-step process: translation, counter-translation and drafting of a provisional version based on the comparison of the translations against the original version. This was subsequently pre-tested on 25 patients at a health centre not participating in the study to make sure the issues were understood correctly by potential study participants and to check the interview time. After modifications, a new version was tested on 15 patients in another health centre not participating in the study. The final questions are presented in Box 1. Responses were categorised using a Likert scale (from 1 to 5): (1) not at all, (2) no, (3) I do not know (neither agree nor disagree), (4) somewhat and (5) a lot. After the last consultation, a self-administered questionnaire was completed by the care providers who conducted the consultations.
(See Box 1 in the Supplementary Files)
This questionnaire was also based on the PPS, but took the perspective of the provider (see Box 2). Testing and translation of this questionnaire followed the same procedures as the version for patients.
(See Box 2 in the Supplementary Files)
4. Data analysis
Patient demographics (age, sex, education level), characteristics of the respondent - i.e. patients or their caretaker/companion) and duration of consultation were analysed using descriptive statistics. Question 7 was excluded from the analyses because 79% of patients did not answer this question, probably because it was seen as either not applicable or as a repeat of question 4. In order to evaluate the internal consistency of the 6 PPS questions (Q1-6), the McDonald’s total omega and hierarchical omega coefficients were calculated. A global participation score was created by summing all scores of the PPS items. A bivariate analysis between the variables of the health centres, providers, MH+ or MH- clusters on the one hand, and the variable 'participation score' on the other hand was then conducted. Multiple regression was conducted with the continuous variable 'participation score' as the dependent variable, and the independent variables that were significantly associated with the dependent variable. To control for the fact that patients were clustered by different providers and to account for non-normality, we used a 'Generalised Estimated Equations (GEE)' model for the analyses.
The answers to the questions were also analysed separately. Since participants’ scores were not distributed normally, they were regrouped in a dichotomous variable with the following categories: "agree" (initial answers options 4 and 5) and "disagree" (initial answers options 1 and 2). Answer option 3 "I do not know (unresolved opinion)" was considered neutral and was ignored in this part of the analysis. The proportion of participants who “agreed” to the different questions were compared between MH+ and MH- health centres, among the different providers, and between patients and providers. These proportions are also graphically represented to allow for a visual comparison of potential differences as recommended by Hirsh et al. (32).
The effect of integration of mental health was assessed by multi-level logistic regression for each of the questions (except for question 8). Demographic characteristics of patients that may have influenced patient-provider interaction, such as age, sex, and level of education were included as independent variables in this model. Providers were considered as a random effect. The model did not include the identity of the respondent (i.e. the patient or their caretaker/companion; the latter in case of consultations of under-fives) since no significant association was found between this particular feature and the responses. We could not make an analysis of question 6 of the PPS because the model did not converge, even when increasing the number of iterations and changing the model’s starting value. The intra-class correlation coefficient was calculated to estimate the extent to which the random variable or the care provider characteristics (age, gender, education) explained the variance of the calculated results. The analyses were carried out using SPSS and R software. For the multilevel analyses, the "glmer" function of the "lme4" package (33) in R and the GEE function in SPSS were used.
The study was approved by the National Commission on Ethics of Health Research (CNERS) of Guinea (N° 010/CNRS/17). The participants were informed about the purpose and the potential risks of the study before the start of the interview and freely agreed to participate. All data collected can be accessed by contacting the first author and principal investigator of the research.